cool to learn @allen_ai are using @vespaengine! and binarized embeddings are amazing for cost/perf, this is a great model.

RT @vespaengine: Perplexity builds AI Search on Blog post below 👇.

RT @vespaengine: When we showed we were 6 to 13 times more efficient than Elastic, our team thought we should do better. Now we're bringin….

And shouldn't that be the goal for agents too - that they can find and start using the services needed to accomplish goals, without human intervention? Can we do for the world what Jeff Bezos did for Amazon?.

I wish the Model Context Protocol would be extended to allow automated signup and payment for APIs. Then I can pretend to be an AI agent and never see "call us for a quote" ever again . .

Shocked at the potential perf improvement from using CBOR over JSON. No more cpu wasted escaping strings and stringifying floats. Good things coming soon I hope.

RT @vespaengine: A tutorial for semantic search in video using @twelve_labs and Vespa:

I keep finding new things to improve because the AI is messing up the config in new and exciting ways, like making our embedding inference code throw ArrayIndexOutOfBoundsException.

Second day of coding a small python project with Windsurf. It feels more like watching TV than writing code. Some times it's magical, some times just infuriating - when I said add a progress bar, it also decided to swap out my dataset?.

Evaluate your retrieval system! It's the best way to drive improvements in quality, and identify the right tradeoffs to cut latency and cost. Super happy to see that it's now easier than ever to do, built right into our Vespa python api with good examples!.

RT @tian_yi_wang: 100% agree. An information retrieval system that scales with intelligence is the future.

RT @JacquesThibs: New Claude model just dropped

RT @dainius_jocas: A new blog post on how stress-free your life is when you run your search on @vespaengine even at a billion-scale!.

This unlocks so many new use cases. Embed and retrieve from all your stuff, super fast, with vastly lower memory and storage cost.

Just landed in London to talk about and RAG at #AICamp tomorrow: Event is fully booked but there's a waiting list, and I have a few openings in my schedule. DM if you want to talk about Vespa!.

Coming back down from the mountains at the end of easter. Feat: passing family and friends due to my better ski technique (aka higher body weight)

Number of the day: One average serving of spaghetti is (at my house) approximately 40 meters, or 130 feet

Oh and isn't California supposed to be warm and dry.

In the bay area talking to people about AI and information retrieval with Vespa. So many good conversations. How would I ever get anything done if I lived here.

And I retweeted the wrong thing so here's the link